Patchwork [net-next,1/1] e1000e: Expose MDI-X state via sysfs

login
register
mail settings
Submitter Chaitanya Lala
Date May 19, 2009, 9:05 p.m.
Message ID <20090519210553.GA2491@clala-laptop>
Download mbox | patch
Permalink /patch/27426/
State Not Applicable
Delegated to: David Miller
Headers show

Comments

Chaitanya Lala - May 19, 2009, 9:05 p.m.
While debugging network connectivity problems, it is often helpful
to report the MDI-X state. The is_mdix variable holds the current
state which we expose on a per-interface basis as a sysfs attribute.
We use sysfs over methods such as netlink due to the convenience of
reading a file (using the cat command) as opposed to connecting to a
netlink socket. If we use a fiber PHY then is_mdix will always be zero
as the mdi-x feature only applies to copper PHYs.

Signed-off-by: Chaitanya Lala <clala@riverbed.com>
Signed-off-by: Arthur Jones <ajones@riverbed.com>
---
 drivers/net/e1000e/netdev.c |   35 +++++++++++++++++++++++++++++++++++
 1 files changed, 35 insertions(+), 0 deletions(-)
Jeff Kirsher - May 19, 2009, 9:55 p.m.
On Tue, May 19, 2009 at 2:05 PM, Chaitanya Lala <clala@riverbed.com> wrote:
> While debugging network connectivity problems, it is often helpful
> to report the MDI-X state. The is_mdix variable holds the current
> state which we expose on a per-interface basis as a sysfs attribute.
> We use sysfs over methods such as netlink due to the convenience of
> reading a file (using the cat command) as opposed to connecting to a
> netlink socket. If we use a fiber PHY then is_mdix will always be zero
> as the mdi-x feature only applies to copper PHYs.
>
> Signed-off-by: Chaitanya Lala <clala@riverbed.com>
> Signed-off-by: Arthur Jones <ajones@riverbed.com>
> ---

NAK.  We do not want to be adding sysfs entries for every little piece
of information in the driver.  Instead, I would suggest looking at
enhancing existing tools like ethtool to get that sort of information
in a more generic way which is not driver specific.

>  drivers/net/e1000e/netdev.c |   35 +++++++++++++++++++++++++++++++++++
>  1 files changed, 35 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
> index ccaaee0..1c131b6 100644
> --- a/drivers/net/e1000e/netdev.c
> +++ b/drivers/net/e1000e/netdev.c
> @@ -4765,6 +4765,32 @@ static const struct net_device_ops e1000e_netdev_ops = {
>  #endif
>  };
>
> +static ssize_t e1000e_show_is_mdix(struct device *dev,
> +    struct device_attribute *attr, char *buf)
> +{
> +       struct net_device *netdev = container_of(dev, struct net_device, dev);
> +       struct e1000_adapter *adapter;
> +       struct e1000_hw *hw;
> +       ssize_t ret = -EINVAL;
> +
> +       if (!buf)
> +               goto err;
> +
> +       read_lock(&dev_base_lock);
> +       if (NETREG_REGISTERED == netdev->reg_state) {
> +               adapter = netdev_priv(netdev);
> +               hw = &adapter->hw;
> +               ret = sprintf(buf, "%d\n", (hw->phy.is_mdix ? 1 : 0));
> +       }
> +       read_unlock(&dev_base_lock);
> +err:
> +       return ret;
> +}
> +
> +/* Export attributes for the device */
> +static struct device_attribute device_attr_is_mdix =
> +    __ATTR(is_mdix, S_IRUGO, e1000e_show_is_mdix, NULL);
> +
>  /**
>  * e1000_probe - Device Initialization Routine
>  * @pdev: PCI device information struct
> @@ -5046,6 +5072,10 @@ static int __devinit e1000_probe(struct pci_dev *pdev,
>        if (err)
>                goto err_register;
>
> +       err = device_create_file(&netdev->dev, &device_attr_is_mdix);
> +       if (err)
> +               goto err_sys_attr;
> +
>        /* carrier off reporting is important to ethtool even BEFORE open */
>        netif_carrier_off(netdev);
>
> @@ -5053,6 +5083,9 @@ static int __devinit e1000_probe(struct pci_dev *pdev,
>
>        return 0;
>
> +err_sys_attr:
> +       unregister_netdev(netdev);
> +
>  err_register:
>        if (!(adapter->flags & FLAG_HAS_AMT))
>                e1000_release_hw_control(adapter);
> @@ -5105,6 +5138,8 @@ static void __devexit e1000_remove(struct pci_dev *pdev)
>
>        flush_scheduled_work();
>
> +       device_remove_file(&netdev->dev, &device_attr_is_mdix);
> +
>        /*
>         * Release control of h/w to f/w.  If f/w is AMT enabled, this
>         * would have already happened in close and is redundant.
> --
> 1.6.0.4
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
Chaitanya Lala - May 19, 2009, 10:22 p.m.
Jeff Kirsher wrote:
> On Tue, May 19, 2009 at 2:05 PM, Chaitanya Lala <clala@riverbed.com> wrote:
>   
>> While debugging network connectivity problems, it is often helpful
>> to report the MDI-X state. The is_mdix variable holds the current
>> state which we expose on a per-interface basis as a sysfs attribute.
>> We use sysfs over methods such as netlink due to the convenience of
>> reading a file (using the cat command) as opposed to connecting to a
>> netlink socket. If we use a fiber PHY then is_mdix will always be zero
>> as the mdi-x feature only applies to copper PHYs.
>>
>> Signed-off-by: Chaitanya Lala <clala@riverbed.com>
>> Signed-off-by: Arthur Jones <ajones@riverbed.com>
>> ---
>>     
>
> NAK.  We do not want to be adding sysfs entries for every little piece
> of information in the driver.  Instead, I would suggest looking at
> enhancing existing tools like ethtool to get that sort of information
> in a more generic way which is not driver specific.
>
>   
The MDI-X setting is a non-standard piece of information & every driver 
may or may-not have it. But still this is an important de-bugging tool & 
we would want to use this information from the drivers that do provide 
this facility.  So what would be a standard way to do this via something 
like ethtool ? Any pointers would be helpful.

Thanks,
Chaitanya
>>  drivers/net/e1000e/netdev.c |   35 +++++++++++++++++++++++++++++++++++
>>  1 files changed, 35 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
>> index ccaaee0..1c131b6 100644
>> --- a/drivers/net/e1000e/netdev.c
>> +++ b/drivers/net/e1000e/netdev.c
>> @@ -4765,6 +4765,32 @@ static const struct net_device_ops e1000e_netdev_ops = {
>>  #endif
>>  };
>>
>> +static ssize_t e1000e_show_is_mdix(struct device *dev,
>> +    struct device_attribute *attr, char *buf)
>> +{
>> +       struct net_device *netdev = container_of(dev, struct net_device, dev);
>> +       struct e1000_adapter *adapter;
>> +       struct e1000_hw *hw;
>> +       ssize_t ret = -EINVAL;
>> +
>> +       if (!buf)
>> +               goto err;
>> +
>> +       read_lock(&dev_base_lock);
>> +       if (NETREG_REGISTERED == netdev->reg_state) {
>> +               adapter = netdev_priv(netdev);
>> +               hw = &adapter->hw;
>> +               ret = sprintf(buf, "%d\n", (hw->phy.is_mdix ? 1 : 0));
>> +       }
>> +       read_unlock(&dev_base_lock);
>> +err:
>> +       return ret;
>> +}
>> +
>> +/* Export attributes for the device */
>> +static struct device_attribute device_attr_is_mdix =
>> +    __ATTR(is_mdix, S_IRUGO, e1000e_show_is_mdix, NULL);
>> +
>>  /**
>>  * e1000_probe - Device Initialization Routine
>>  * @pdev: PCI device information struct
>> @@ -5046,6 +5072,10 @@ static int __devinit e1000_probe(struct pci_dev *pdev,
>>        if (err)
>>                goto err_register;
>>
>> +       err = device_create_file(&netdev->dev, &device_attr_is_mdix);
>> +       if (err)
>> +               goto err_sys_attr;
>> +
>>        /* carrier off reporting is important to ethtool even BEFORE open */
>>        netif_carrier_off(netdev);
>>
>> @@ -5053,6 +5083,9 @@ static int __devinit e1000_probe(struct pci_dev *pdev,
>>
>>        return 0;
>>
>> +err_sys_attr:
>> +       unregister_netdev(netdev);
>> +
>>  err_register:
>>        if (!(adapter->flags & FLAG_HAS_AMT))
>>                e1000_release_hw_control(adapter);
>> @@ -5105,6 +5138,8 @@ static void __devexit e1000_remove(struct pci_dev *pdev)
>>
>>        flush_scheduled_work();
>>
>> +       device_remove_file(&netdev->dev, &device_attr_is_mdix);
>> +
>>        /*
>>         * Release control of h/w to f/w.  If f/w is AMT enabled, this
>>         * would have already happened in close and is redundant.
>> --
>> 1.6.0.4
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe netdev" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>>     
>
>
>
>   

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jesse Brandeburg - May 19, 2009, 11:45 p.m.
Chaitanya Lala wrote:
> Jeff Kirsher wrote:
>> On Tue, May 19, 2009 at 2:05 PM, Chaitanya Lala <clala@riverbed.com>
>> wrote:
>>
>>> While debugging network connectivity problems, it is often helpful
>>> to report the MDI-X state. The is_mdix variable holds the current
>>> state which we expose on a per-interface basis as a sysfs attribute.
>>> We use sysfs over methods such as netlink due to the convenience of
>>> reading a file (using the cat command) as opposed to connecting to a
>>> netlink socket. If we use a fiber PHY then is_mdix will always be
>>> zero
>>> as the mdi-x feature only applies to copper PHYs.
>>>
>>> Signed-off-by: Chaitanya Lala <clala@riverbed.com>
>>> Signed-off-by: Arthur Jones <ajones@riverbed.com>
>>> ---
>>>
>>
>> NAK.  We do not want to be adding sysfs entries for every little
>> piece
>> of information in the driver.  Instead, I would suggest looking at
>> enhancing existing tools like ethtool to get that sort of information
>> in a more generic way which is not driver specific.
>>
>>
> The MDI-X setting is a non-standard piece of information & every
> driver may or may-not have it. But still this is an important
> de-bugging tool & we would want to use this information from the
> drivers that do provide this facility.  So what would be a standard
> way to do this via something like ethtool ? Any pointers would be
> helpful.

how about (psuedo diff)

        Supported ports: [ TP ]
        Supported link modes:   10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
                                1000baseT/Full
        Supports auto-negotiation: Yes
        Advertised link modes:  10baseT/Half 10baseT/Full
                                100baseT/Half 100baseT/Full
                                1000baseT/Full
        Advertised auto-negotiation: Yes
        Speed: 1000Mb/s
        Duplex: Full
        Port: Twisted Pair
        PHYAD: 1
        Transceiver: internal
        Auto-negotiation: on
+       MDI-X: on
        Supports Wake-on: umbg
        Wake-on: g
        Current message level: 0x00000003 (3)
        Link detected: yes

where possible MDI-X values are (on|off)

This way drivers can optionally implement the feature in their 
get_settings and set_settings handlers.
Chaitanya Lala - May 20, 2009, 5:14 p.m.
Brandeburg, Jesse wrote:
> Chaitanya Lala wrote:
>   
>> Jeff Kirsher wrote:
>>     
>>> On Tue, May 19, 2009 at 2:05 PM, Chaitanya Lala <clala@riverbed.com>
>>> wrote:
>>>
>>>       
>>>> While debugging network connectivity problems, it is often helpful
>>>> to report the MDI-X state. The is_mdix variable holds the current
>>>> state which we expose on a per-interface basis as a sysfs attribute.
>>>> We use sysfs over methods such as netlink due to the convenience of
>>>> reading a file (using the cat command) as opposed to connecting to a
>>>> netlink socket. If we use a fiber PHY then is_mdix will always be
>>>> zero
>>>> as the mdi-x feature only applies to copper PHYs.
>>>>
>>>> Signed-off-by: Chaitanya Lala <clala@riverbed.com>
>>>> Signed-off-by: Arthur Jones <ajones@riverbed.com>
>>>> ---
>>>>
>>>>         
>>> NAK.  We do not want to be adding sysfs entries for every little
>>> piece
>>> of information in the driver.  Instead, I would suggest looking at
>>> enhancing existing tools like ethtool to get that sort of information
>>> in a more generic way which is not driver specific.
>>>
>>>
>>>       
>> The MDI-X setting is a non-standard piece of information & every
>> driver may or may-not have it. But still this is an important
>> de-bugging tool & we would want to use this information from the
>> drivers that do provide this facility.  So what would be a standard
>> way to do this via something like ethtool ? Any pointers would be
>> helpful.
>>     
>
> how about (psuedo diff)
>
>         Supported ports: [ TP ]
>         Supported link modes:   10baseT/Half 10baseT/Full
>                                 100baseT/Half 100baseT/Full
>                                 1000baseT/Full
>         Supports auto-negotiation: Yes
>         Advertised link modes:  10baseT/Half 10baseT/Full
>                                 100baseT/Half 100baseT/Full
>                                 1000baseT/Full
>         Advertised auto-negotiation: Yes
>         Speed: 1000Mb/s
>         Duplex: Full
>         Port: Twisted Pair
>         PHYAD: 1
>         Transceiver: internal
>         Auto-negotiation: on
> +       MDI-X: on
>         Supports Wake-on: umbg
>         Wake-on: g
>         Current message level: 0x00000003 (3)
>         Link detected: yes
>
> where possible MDI-X values are (on|off)
>
> This way drivers can optionally implement the feature in their 
> get_settings and set_settings handlers.
>   
Thanks for the possible solution. I will work on it and get back.
Thanks,
Chaitanya
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
index ccaaee0..1c131b6 100644
--- a/drivers/net/e1000e/netdev.c
+++ b/drivers/net/e1000e/netdev.c
@@ -4765,6 +4765,32 @@  static const struct net_device_ops e1000e_netdev_ops = {
 #endif
 };
 
+static ssize_t e1000e_show_is_mdix(struct device *dev,
+    struct device_attribute *attr, char *buf)
+{
+	struct net_device *netdev = container_of(dev, struct net_device, dev);
+	struct e1000_adapter *adapter;
+	struct e1000_hw *hw;
+	ssize_t ret = -EINVAL;
+
+	if (!buf)
+		goto err;
+
+	read_lock(&dev_base_lock);
+	if (NETREG_REGISTERED == netdev->reg_state) {
+		adapter = netdev_priv(netdev);
+		hw = &adapter->hw;
+		ret = sprintf(buf, "%d\n", (hw->phy.is_mdix ? 1 : 0));
+	}
+	read_unlock(&dev_base_lock);
+err:
+	return ret;
+}
+
+/* Export attributes for the device */
+static struct device_attribute device_attr_is_mdix =
+    __ATTR(is_mdix, S_IRUGO, e1000e_show_is_mdix, NULL);
+
 /**
  * e1000_probe - Device Initialization Routine
  * @pdev: PCI device information struct
@@ -5046,6 +5072,10 @@  static int __devinit e1000_probe(struct pci_dev *pdev,
 	if (err)
 		goto err_register;
 
+	err = device_create_file(&netdev->dev, &device_attr_is_mdix);
+	if (err)
+		goto err_sys_attr;
+
 	/* carrier off reporting is important to ethtool even BEFORE open */
 	netif_carrier_off(netdev);
 
@@ -5053,6 +5083,9 @@  static int __devinit e1000_probe(struct pci_dev *pdev,
 
 	return 0;
 
+err_sys_attr:
+	unregister_netdev(netdev);
+
 err_register:
 	if (!(adapter->flags & FLAG_HAS_AMT))
 		e1000_release_hw_control(adapter);
@@ -5105,6 +5138,8 @@  static void __devexit e1000_remove(struct pci_dev *pdev)
 
 	flush_scheduled_work();
 
+	device_remove_file(&netdev->dev, &device_attr_is_mdix);
+
 	/*
 	 * Release control of h/w to f/w.  If f/w is AMT enabled, this
 	 * would have already happened in close and is redundant.